Intraoperative fluorescence imaging with aminolevulinic acid detects grossly occult breast cancer: a phase II randomized controlled trial

Breast cancer, the most prevalent cancer in women, is often treated by breast-conserving surgery (BCS) [1, 2] which aims to completely excise the tumor with clear margins while preserving the maximum amount of healthy tissue. Cancer visualization under standard white light (WL) operating room conditions is difficult due to low cancer-to-normal tissue contrast, resulting in positive margins which are related to greater risk of local recurrence (LR) [3‐6], inferior outcomes (complications, stress, poor cosmesis) [7], adjuvant delay [8], increased healthcare costs [9‐11], and lower disease-specific survival [12]. Furthermore, incomplete resection requires reoperation in 20–70% of patients [2, 6, 13‐20]. Thus, reducing positive margin rates is a major goal in BCS [21] and is an internationally recognized quality indicator for treatment [7, 22].

Currently, BCS best practice uses WL visualization, palpation, specimen radiography, and intraoperative histopathology to guide resection. These techniques are lengthy (~ 20 min) [23], limited by inaccurate co-localization of positive margins on the excised tissue to the surgical bed [24] and variably impact outcomes [23]. Adaptation of standard medical imaging technologies for operating room use (MRI [25], ultrasound [25, 26], PET/CT [25]), and emerging intraoperative tumor detection technologies for BCS, both non-optical and optically enabled, are either at the preclinical stage or are not widely adopted due to practical limitations. Thus, there is a clinical need for an alternative, safe, practical, cost-sensitive, and real-time intraoperative imaging technology for surgeons and pathologists to visualize occult malignancy in excised specimens and surgical cavities during index BCS.

Fluorescence imaging using an intraoperative instrument plus an exogenous tumor-specific imaging agent may fulfill this need by enhancing tumor tissue contrast and facilitating the detection of grossly occult disease. Intraoperative fluorescence imaging for detection of carcinoma has been demonstrated in clinical trials for several tumor types using visible and near-infrared (NIR) contrast agents [27‐30]. However, the instrumentation involves large, costly cart-based systems that are impractical for BCS since they do not fulfill the surgeon’s need to interrogate both the surgical specimen and cavity. To address this, we developed a handheld fluorescence imaging device (Portable Real-time Optical Detection Identification and Guide for Intervention, PRODIGI) for real-time intraoperative fluorescence imaging of excised breast specimens and the surgical cavity. PRODIGI is clinically safe and has demonstrated clinical utility in other medical applications [31‐33]. The device combines consumer-grade imaging sensor technology with miniature light-emitting diodes (LEDs) and multiband optical filtering to capture high-resolution WL and fluorescence digital images and videos. PRODIGI’s multiband optical filter allows for imaging of porphyrins, including protoporphyrin IX (PpIX), which fluoresces red (peak emission at 635 nm wavelength) [34, 35] when excited by violet-blue light (~ 400–410 nm). PpIX is a metabolite of the prodrug 5-aminolevulinic acid (5-ALA), an endogenous non-protein amino acid involved in heme biosynthesis in mammalian cells. Unlike contrast agents that are receptor-targeted or enzyme-activated, PpIX is not affected by suboptimal biodistribution and tumor penetration, heterogenous target receptor/enzyme expression [29], and high background fluorescence [29, 36].

Delivered systemically, 5-ALA is taken up by cells throughout the body where it is converted into heme. In aberrant cells, such as cancer cells, defects in heme biosynthesis cause accumulation of PpIX [35, 37‐42], enabling real-time visualization. 5-ALA is widely known for its clinical use in photodynamic diagnosis [43‐55] and therapy [34, 49, 56, 57] of premalignant and malignant disease. Large clinical trials have demonstrated the safety and clinical utility of 5-ALA-based fluorescence image-guided surgery (FIGS) for malignant glioma [58‐61], which led to the approval of oral 5-ALA hydrochloride (HCl) for FIGS of high-grade glioma in approximately 40 countries, including the USA [62]. Other clinical studies support 5-ALA for the visualization of malignant tissue in the bladder [44], gastrointestinal tract [50], oral cavity [51, 52], lung [47, 55], and female genital tract [49, 53, 54]. 5-ALA has a well-established safety profile, which is highlighted in the Gleolan Prescribing Information for glioma patients where pyrexia, hypotension, nausea, and vomiting are reported to have occurred in > 1% of patient’s in the week following surgery and chills, photosensitivity reaction, solar dermatitis, hypotension, abnormal liver function test, and diarrhea are reported to have occurred in < 1% of patients in the 6 weeks after surgery [62]. Only one other clinical study (in 2001) explored the utility of 5-ALA (40 mg/kg bodyweight (BW)) for detecting breast cancers in a non-randomized study of 16 patients, although the results were exploratory and limited to imaging surgical specimens only [48]. Since then, no follow-up studies on the application of 5-ALA in breast cancer imaging have been published.

In this report, we present the results of a single-center non-interventional Phase II randomized controlled trial (RCT) designed to characterize the imaging performance of our handheld imaging device with two doses (15 and 30 mg/kg BW) of 5-ALA HCl versus no tumor contrast in patients with invasive breast cancer undergoing lumpectomy or mastectomy. Two doses (15 or 30 mg/kg BW) of 5-ALA HCl were administered to evaluate the minimum dosage for sufficient tumor-to-normal tissue fluorescence contrast obtained using the PRODIGI device. The doses selected are “on either side” of the FDA-approved dose for glioma (20 mg/kg) [62] and are based on previous clinical studies validating safety at ≤ 60 mg/kg [63‐65], tumor enhancement as low as 10 mg/kg [59, 60, 64], and improved tumor-to-normal tissue contrast with increased doses [60, 64]. The primary objective was to determine measures of diagnostic accuracy (positive and negative predictive values, sensitivity, specificity, and diagnostic odds ratio) of visualization of breast tumors in surgical specimens using PRODIGI in combination with 5-ALA HCl, with ex vivo biopsy-based histology as the gold standard. The secondary objective was to evaluate the feasibility and safety of the PRODIGI device for intraoperative imaging of the surgical cavity following resection. Finally, we confirm the safety and technical feasibility and discuss the future clinical adoptability of PRODIGI plus 5-ALA HCl to improve tumor visualization during BCS.

Surgical margins are a significant challenge in the treatment of solid cancers [11]. During this study the Society of Surgical Oncology/American Society Radiation Oncology/American Society of Clinical Oncology released joint guidelines for BCS recommending a standard definition of a positive margin based on clinical evidence [6, 70]. The guidelines define positive margins as “ink on tumor” for invasive breast cancer [6] and cancer within 2 mm of the inked margin for DCIS [70]. Surgeons disagree regarding the recommendations [19, 71, 72] and the effect of the guidelines on re-excision rates varies [71, 73‐79]. Many studies still report suboptimal re-excision rates [71, 74, 76, 79‐81], above the internationally accepted target of 10% [7, 22], reinforcing the need for new imaging methods that allow surgeons and pathologists to visualize grossly occult disease on the lumpectomy and in the surgical cavity during index surgery.

5-ALA was well tolerated and enabled visualization of grossly obvious tumors and occult disease across all tumor grades (I-III) and types, including IDC and ILC with and without in situ disease. Both doses tested were lower than the only dose previously reported for clinical use in breast cancer (40 mg/kg) [48] and consistently induced bright red cancer fluorescence (contrasted against normal tissue AF) detected by the device.

PRODIGI was easy to use and practical due to its handheld form factor (e.g., non-cart based), which overcomes the large footprint limitation of existing closed- and wide-field fluorescence surgical imaging systems. The device was safely integrated into the standard surgical workflow in both the operating room and the pathology suite without affecting standard specimen processing procedures. PpIX fluorescence imaging with PRODIGI was feasible in the presence of blue dye, which is commonly used in sentinel node mapping. Although there is a potential for the blue dye to absorb light in the red wavelength range, PpIX was visualized in tissues stained with blue dye. PRODIGI imaging within a small lumpectomy cavity was impeded by the form factor of the prototype which led to shadow artefacts due to the position of the LEDs. Additionally, although a standardized imaging distance of ~ 10 cm was maintained when imaging excised tissue on the bench, achieving consistent distance for intracavity imaging was challenging. The next-generation device, currently being developed by our group, is anticipated to overcome these challenges by improving the optical power density output, re-configuring the optical components to maximize cavity and specimen surface illumination, including range finder technology to ensure a standardized working distance is maintained for optimal imaging and re-designing the form factor. For the free-hand imaging modality of the PRODIGI device, there is a potential for images taken to be blurry. We acknowledge that some of our images (Fig. 2A(b) and Fig. 2B(d)) are blurry and were not frames from a video. A blurry image, however, did not detract from real-time visualization of PpIX fluorescence. The next-generation device will include a tap-to-focus feature. A multicenter Phase III interventional imaging RCT will use this optimized device to evaluate the effectiveness of 5-ALA-induced PpIX fluorescence image-guided margin assessment and resection during BCS. Future studies may also evaluate fluorescence imaging during gross examination of specimens to facilitate rapid, targeted specimen sectioning and identification of areas for touch-prep cytology and frozen-section analysis, to improve rapid decisions about additional tissue excision and location of positive margins in the surgical cavity.

Predictive values depend on disease prevalence, with a higher disease prevalence associated with higher PPV and lower NPV, and vice versa [82]. Analysis was, therefore, stratified by biopsy location as the prevalence of cancer cells is higher inside versus outside the tumor border. In our study, high PPV inside and high NPV outside the tumor border were observed for both 5-ALA doses, indicating that PRODIGI functioned as expected. However, sensitivity and specificity are generally accepted to not be influenced by differences in disease prevalence [83]. Thus, while these calculations are possible with biopsies from inside and outside the tumor border, more meaningful sensitivity and specificity are derived without stratification by biopsy location. When compared to MarginProbe, which uses non-imaging radiofrequency signals to alert surgeons about positive margins [84‐86], our platform may be less likely to result in removal of cancer-negative tissues. Moreover, our approach is more sensitive and comparably specific to conventional BCS [86].

False positives and negatives were seen in both 5-ALA dose groups inside and outside the demarcated tumor border. False positives were only observed in 5-ALA patients (outside the tumor border), and not in controls, indicating likely ALA-induced PpIX production in non-malignant cells. The majority of false positive biopsies (5/9), examined histologically, demonstrated non-malignant proliferative changes (hyperplasia, columnar cell change, atypical ductal hyperplasia, sclerosing adenosis, and fibrocystic change) with infiltrating immune cells present in one biopsy. If we consider the detection of these “abnormal” proliferative changes as a desirable outcome of the test, the PPV to detect abnormal breast tissue (including malignant and non-malignant proliferative changes) increases to 89% and 70% for the low- and high-dose groups, respectively. Other studies using 5-ALA have observed the accumulation of PpIX fluorescence in premalignant diseases [57], in areas with immune cell infiltration [87, 88] and in benign disease, such as Grade 1 meningiomas [89]. Removing non-malignant tissues that are associated with increased breast cancer risk [90] may outweigh the potential effect on cosmesis during BSC, minimizing the clinical impact of false positives. Unlike other non-targeted approaches, such as circumferential cavity shaving [81], fluorescence imaging would conceivably allow for more spatially targeted revision of margins, reducing unnecessary removal of additional tissue which is desirable in optimizing cosmesis. Nevertheless, additional studies are needed to understand how 5-ALA behaves in benign and malignant breast tissues to improve their differentiation during intraoperative fluorescence imaging.

Eleven of the 14 false negative biopsies were collected from inside the tumor border, the majority of which (8/11) came from primary tumors that had observable red fluorescence elsewhere inside the demarcated tumor or in an area contiguous with the demarcated tumor on an adjacent slice, suggesting heterogeneity in biological factors that may influence 5-ALA uptake and the production, accumulation, and/or detection of PpIX in the primary tumor. For example, differences in tissue composition can affect optical properties and light transport in tissue which can affect the detection of PpIX. Moreover, endogenous fluorophores that emit high-intensity green AF (collagen, flavin adenine dinucleotide-FAD, reduced nicotinamide adenine dinucleotide-NADH) [91] may “mask” PpIX fluorescence, especially in small cancer foci.

Multifocal disease, associated with poorer disease-free survival [92], was also visualized in serially sectioned specimens. PpIX fluorescence revealed lobulated tumor borders and heterogenous tumor composition, which was consistent with patients’ preoperative imaging/pathology but not appreciable by palpation. While some specimens had uniform bright red fluorescence tumors, others had heterogeneous (multifocal) bright red fluorescence foci with green connective tissue interspersed throughout the tumor, which is consistent with carcinoma-associated fibrosis [93]. We observed spatial heterogeneity in the density of fluorescent connective tissues in the surgical specimens during macroscopic examination with some specimens being homogeneously bright green and others showing a more mottled appearance. Moreover, increased microvasculature may preferentially absorb the 405 nm excitation light and emitted PpIX fluorescence. These findings may aid decision-making for oncoplastic surgical techniques [94] and identify patients with residual or multicentric disease beyond the primary tumor.

Limitations of the study include its single-center design, variability in the time to imaging, restriction on the total number of biopsies collected per patient, and a focus on visualization of carcinoma inside the serially sectioned specimen and not on the specimen margins. PMCC draws patients from various regional community practices representative of patients seen throughout Canada and the U.S. However, clinical practice by breast surgeons (A.M.E., W.L.L.) in our academic setting may vary from community practices [95]. Variability in the imaging time relative to administration of 5-ALA between patients is unavoidable in a surgical intervention study at a high-volume center. Nevertheless, we aimed to administer 5-ALA 2-4 h before anesthesia, to enable imaging between 2.5 and 5 h post-administration, which is clinically relevant for maximal PpIX signal [96]. The institution restricted the number and size of biopsies collected from inside and outside the demarcated tumor border. To maximize the number of PpIX red fluorescence areas analyzed histologically, biopsies were collected under fluorescence image guidance, possibly leading to a sampling selection bias. However, given the restrictions on biopsy collection, it would be difficult to collect enough biopsies without fluorescence image guidance to calculate sensitivity outside the tumor border. Lastly, future studies will assess the accuracy of 5-ALA-induced fluorescence visualization for the detection of positive margins during BCS.

Intraoperative fluorescence imaging is an emerging technique with significant potential to improve clinical outcomes in breast cancer patients. The superficial (~ 2 mm [56]) penetration of PRODIGI’s 405 nm excitation light is more suitable for intraoperative margin assessment compared to NIR fluorescence imaging methods, which can image to depths of several centimeters [27, 28, 97, 98] and are, therefore, more likely to result in false positive margin readings. Unlike other non-image-based methods, including the FDA-approved MarginProbe [86], our approach produces a real-time color image of the anatomical location and extent of carcinoma with sub-millimeter resolution. The differing emission spectra of normal tissue AF and PpIX are visually distinguishable, providing excellent fluorescence contrast of the cancers against background healthy tissues and simplifying fluorescence image interpretation for the operator [99]. This facilitates real-time decision-making by guiding surgeons to potential areas of microscopic disease within a macroscopic setting [99], and enabling treatment plan adjustments, which is not possible with current preoperative imaging modalities.

Our study also contributes to the growing clinical evidence of 5-ALA for intraoperative visualization of clinically occult cancer beyond neurosurgery where this approach is becoming the standard of care [43]. The versatile design of the novel handheld fluorescence imaging device could help to translate this approach to other cancer surgeries where margin assessment and image-guided resection are valuable. Future clinical adoption will be predicated on safe and effective device and contrast agent combinations that are easily integrated into the operating room workflow, meet the user’s needs (wireless one-handed use), and are cost-sensitive.